In this case the servers use a virtual (alias) IP address (called Service IP), clients access the service provided by the servers by targeting this Service IP. The Service IP is carried by the active node only. If the node providing the service (that is, the master node) fails, the Service IP is taken over by the slave node. As all clients send requests towards the Service IP, they are not aware of which device that address belongs to, and do not notice any difference when a takeover occurs.
When the cluster relocates the Service IP to the other node it sends a gratuitous arp request message to the whole network informing the clients that the Service IP belongs now to a different node. As a result, the clients flush their arp cache and record the new arp address of the Service IP. (A gratuitous ARP request is an AddressResolutionProtocol request packet where the source and destination IP are both set to the IP of the machine issuing the packet and the destination MAC is the broadcast address ff:ff:ff:ff:ff:ff. Ordinarily, no reply packet will occur.)
Service IP takeover is the most frequently used takeover method for PNS clusters.
| Note |
|---|
Some clients may not take over the new Service IP address until the next automatic arp cache flush, which causes certain delay in Service IP transfers in the system. The problem is that the arp cache is refreshed relatively rarely, and it is not possible to notify the clients that they should update their arp cache. |
Published on June 04, 2020
© 2007-2019 BalaSys
Send your comments to support@balasys.hu
